Skip to content

Commit f407f2d

Browse files
committed
AMDGPU: Test a few more cycle counts for mfma ops
1 parent c54064d commit f407f2d

File tree

1 file changed

+74
-12
lines changed

1 file changed

+74
-12
lines changed

llvm/test/tools/llvm-mca/AMDGPU/gfx940-mfma.s

Lines changed: 74 additions & 12 deletions
Original file line numberDiff line numberDiff line change
@@ -1,18 +1,9 @@
11
# RUN: llvm-mca -mtriple=amdgcn -mcpu=gfx940 --timeline --iterations=1 --timeline-max-cycles=0 < %s | FileCheck %s
22

33
# CHECK: Iterations: 1
4-
# CHECK: Instructions: 58
5-
# CHECK: Total Cycles: 543
6-
# CHECK: Total uOps: 58
7-
8-
# CHECK: Resources:
9-
# CHECK: [0] - HWBranch
10-
# CHECK: [1] - HWExport
11-
# CHECK: [2] - HWLGKM
12-
# CHECK: [3] - HWSALU
13-
# CHECK: [4] - HWVALU
14-
# CHECK: [5] - HWVMEM
15-
# CHECK: [6] - HWXDL
4+
# CHECK: Instructions: 78
5+
# CHECK: Total Cycles: 699
6+
# CHECK: Total uOps: 78
167

178
v_mfma_f32_16x16x4_f32 a[0:3], v0, v1, a[2:5]
189
v_mfma_f32_16x16x4_f32 v[0:3], v0, v1, v[2:5]
@@ -101,6 +92,57 @@ v_smfmac_i32_16x16x64_i8 a[10:13], v[2:3], a[4:7], v9
10192
v_smfmac_i32_32x32x32_i8 v[10:25], a[2:3], v[4:7], v10 cbsz:3 abid:1
10293
v_smfmac_i32_32x32x32_i8 a[10:25], v[2:3], a[4:7], v11
10394

95+
v_mfma_f32_16x16x32_bf8_bf8 v[0:3], v[2:3], v[4:5], v[0:3]
96+
v_mfma_f32_16x16x32_bf8_bf8 a[0:3], v[2:3], v[4:5], a[0:3]
97+
98+
v_mfma_f32_16x16x32_bf8_fp8 v[0:3], v[2:3], v[4:5], v[0:3]
99+
v_mfma_f32_16x16x32_bf8_fp8 a[0:3], v[2:3], v[4:5], a[0:3]
100+
101+
v_mfma_f32_16x16x32_fp8_bf8 v[0:3], v[2:3], v[4:5], v[0:3]
102+
v_mfma_f32_16x16x32_fp8_bf8 a[0:3], v[2:3], v[4:5], a[0:3]
103+
104+
v_mfma_f32_16x16x32_fp8_fp8 v[0:3], v[2:3], v[4:5], v[0:3]
105+
v_mfma_f32_16x16x32_fp8_fp8 a[0:3], v[2:3], v[4:5], a[0:3]
106+
107+
v_mfma_f32_32x32x16_bf8_bf8 v[0:15], v[2:3], v[4:5], v[0:15]
108+
v_mfma_f32_32x32x16_fp8_bf8 v[0:15], v[2:3], v[4:5], v[0:15]
109+
v_mfma_f32_32x32x16_bf8_fp8 v[0:15], v[2:3], v[4:5], v[0:15]
110+
v_mfma_f32_32x32x16_fp8_fp8 v[0:15], v[2:3], v[4:5], v[0:15]
111+
112+
v_smfmac_f32_16x16x64_bf8_bf8 v[0:3], a[2:3], v[4:7], v1 cbsz:3 abid:1
113+
v_smfmac_f32_16x16x64_bf8_fp8 v[0:3], a[2:3], v[4:7], v1 cbsz:3 abid:1
114+
v_smfmac_f32_16x16x64_fp8_bf8 v[0:3], a[2:3], v[4:7], v1 cbsz:3 abid:1
115+
v_smfmac_f32_16x16x64_fp8_fp8 v[0:3], a[2:3], v[4:7], v1 cbsz:3 abid:1
116+
117+
v_smfmac_f32_32x32x32_bf8_bf8 v[0:15], v[2:3], v[4:7], v1 cbsz:3 abid:1
118+
v_smfmac_f32_32x32x32_bf8_fp8 v[0:15], v[2:3], v[4:7], v1 cbsz:3 abid:1
119+
v_smfmac_f32_32x32x32_fp8_bf8 v[0:15], v[2:3], v[4:7], v1 cbsz:3 abid:1
120+
v_smfmac_f32_32x32x32_fp8_fp8 v[0:15], v[2:3], v[4:7], v1 cbsz:3 abid:1
121+
122+
# CHECK: Instruction Info:
123+
# CHECK-NEXT:[1]: #uOps
124+
# CHECK-NEXT:[2]: Latency
125+
# CHECK-NEXT:[3]: RThroughput
126+
# CHECK-NEXT:[4]: MayLoad
127+
# CHECK-NEXT:[5]: MayStore
128+
# CHECK-NEXT:[6]: HasSideEffects (U)
129+
130+
# CHECK: [1] [2] [3] [4] [5] [6] Instructions:
131+
132+
# CHECK: 1 8 1.00 U v_mfma_f64_4x4x4_4b_f64 a[0:1], v[0:1], a[2:3], a[2:3]
133+
# CHECK-NEXT: 1 8 1.00 U v_mfma_f64_4x4x4_4b_f64 v[0:1], v[0:1], v[2:3], v[2:3]
134+
# CHECK-NEXT: 1 20 1.00 U v_mfma_f64_16x16x4_f64 a[0:7], v[0:1], v[2:3], a[0:7]
135+
# CHECK-NEXT: 1 20 1.00 U v_mfma_f64_16x16x4_f64 v[0:7], v[0:1], v[2:3], v[0:7]
136+
137+
# CHECK: Resources:
138+
# CHECK: [0] - HWBranch
139+
# CHECK: [1] - HWExport
140+
# CHECK: [2] - HWLGKM
141+
# CHECK: [3] - HWSALU
142+
# CHECK: [4] - HWVALU
143+
# CHECK: [5] - HWVMEM
144+
# CHECK: [6] - HWXDL
145+
104146
# CHECK: [0] [1] [2] [3] [4] [5] [6] Instructions:
105147
# CHECK-NEXT: - - - - - - 8.00 v_mfma_f32_16x16x4_f32 a[0:3], v0, v1, a[2:5]
106148
# CHECK-NEXT: - - - - - - 8.00 v_mfma_f32_16x16x4_f32 v[0:3], v0, v1, v[2:5]
@@ -160,3 +202,23 @@ v_smfmac_i32_32x32x32_i8 a[10:25], v[2:3], a[4:7], v11
160202
# CHECK-NEXT: - - - - - - 4.00 v_smfmac_i32_16x16x64_i8 a[10:13], v[2:3], a[4:7], v9
161203
# CHECK-NEXT: - - - - - - 8.00 v_smfmac_i32_32x32x32_i8 v[10:25], a[2:3], v[4:7], v10 cbsz:3 abid:1
162204
# CHECK-NEXT: - - - - - - 8.00 v_smfmac_i32_32x32x32_i8 a[10:25], v[2:3], a[4:7], v11
205+
# CHECK-NEXT: - - - - - - 4.00 v_mfma_f32_16x16x32_bf8_bf8 v[0:3], v[2:3], v[4:5], v[0:3]
206+
# CHECK-NEXT: - - - - - - 4.00 v_mfma_f32_16x16x32_bf8_bf8 a[0:3], v[2:3], v[4:5], a[0:3]
207+
# CHECK-NEXT: - - - - - - 4.00 v_mfma_f32_16x16x32_bf8_fp8 v[0:3], v[2:3], v[4:5], v[0:3]
208+
# CHECK-NEXT: - - - - - - 4.00 v_mfma_f32_16x16x32_bf8_fp8 a[0:3], v[2:3], v[4:5], a[0:3]
209+
# CHECK-NEXT: - - - - - - 4.00 v_mfma_f32_16x16x32_fp8_bf8 v[0:3], v[2:3], v[4:5], v[0:3]
210+
# CHECK-NEXT: - - - - - - 4.00 v_mfma_f32_16x16x32_fp8_bf8 a[0:3], v[2:3], v[4:5], a[0:3]
211+
# CHECK-NEXT: - - - - - - 4.00 v_mfma_f32_16x16x32_fp8_fp8 v[0:3], v[2:3], v[4:5], v[0:3]
212+
# CHECK-NEXT: - - - - - - 4.00 v_mfma_f32_16x16x32_fp8_fp8 a[0:3], v[2:3], v[4:5], a[0:3]
213+
# CHECK-NEXT: - - - - - - 8.00 v_mfma_f32_32x32x16_bf8_bf8 v[0:15], v[2:3], v[4:5], v[0:15]
214+
# CHECK-NEXT: - - - - - - 8.00 v_mfma_f32_32x32x16_fp8_bf8 v[0:15], v[2:3], v[4:5], v[0:15]
215+
# CHECK-NEXT: - - - - - - 8.00 v_mfma_f32_32x32x16_bf8_fp8 v[0:15], v[2:3], v[4:5], v[0:15]
216+
# CHECK-NEXT: - - - - - - 8.00 v_mfma_f32_32x32x16_fp8_fp8 v[0:15], v[2:3], v[4:5], v[0:15]
217+
# CHECK-NEXT: - - - - - - 4.00 v_smfmac_f32_16x16x64_bf8_bf8 v[0:3], a[2:3], v[4:7], v1 cbsz:3 abid:1
218+
# CHECK-NEXT: - - - - - - 4.00 v_smfmac_f32_16x16x64_bf8_fp8 v[0:3], a[2:3], v[4:7], v1 cbsz:3 abid:1
219+
# CHECK-NEXT: - - - - - - 4.00 v_smfmac_f32_16x16x64_fp8_bf8 v[0:3], a[2:3], v[4:7], v1 cbsz:3 abid:1
220+
# CHECK-NEXT: - - - - - - 4.00 v_smfmac_f32_16x16x64_fp8_fp8 v[0:3], a[2:3], v[4:7], v1 cbsz:3 abid:1
221+
# CHECK-NEXT: - - - - - - 8.00 v_smfmac_f32_32x32x32_bf8_bf8 v[0:15], v[2:3], v[4:7], v1 cbsz:3 abid:1
222+
# CHECK-NEXT: - - - - - - 8.00 v_smfmac_f32_32x32x32_bf8_fp8 v[0:15], v[2:3], v[4:7], v1 cbsz:3 abid:1
223+
# CHECK-NEXT: - - - - - - 8.00 v_smfmac_f32_32x32x32_fp8_bf8 v[0:15], v[2:3], v[4:7], v1 cbsz:3 abid:1
224+
# CHECK-NEXT: - - - - - - 8.00 v_smfmac_f32_32x32x32_fp8_fp8 v[0:15], v[2:3], v[4:7], v1 cbsz:3 abid:1

0 commit comments

Comments
 (0)